Why do we test children on sentences they almost never hear?

If I were to design a comprehension test for children containing sentences such as “Which firewall did the hacker infiltrate?” or “which entrecote did the sous-chef drizzle?” you would probably think I was crazy. The problem with these sentences is that they use rare terminology (from computing and cooking) which young children are unlikely to have heard. It’s obviously crazy to use sentences like this. It’s a good thing that standardised tests contain nice normal garden-variety sentences.

Think again.

Assessments such as the Test of Reception of Grammar (TROG) contain sentences such as “The sheep that the girl looks at is running.” This describes a fairly common situation using simple vocabulary, so what’s wrong with it? The problem is that this sentence is probably just as rare as “Which firewall did the hacker infiltrate?” The issue is the Noun Phrase (the girl). This comes inside a relative clause (that the girl looks at). Relative clauses almost always contain pronouns (e.g. she) because one of their basic functions is to refer back to previously stated information. If you don’t believe me, just read Fox and Thompson (1990).

There is evidence to suggest that the input frequency of particular types of sentences greatly affects children’s ability to process them. For example, Kidd et al. (2007) found that just by changing “the girl” to “she”, a pattern far more frequent in the input, sentences like the above become a lot easier to understand. Another good example of this is questions. Novogrodsky and Friedmann (2011) report that children with Specific Language Impairment find questions such as “which cat was the dog chasing?” particularly difficult to understand. They attribute this finding to a grammatical principle called Relativised Minimality. However, an alternative explanation is that this type of question is extremely rare in the input. In fact, according to my count, only 0.4% of object questions contain both a question phrase such as “which cat”, and a full Noun Phrase such as “the dog.” The exact reason for this low frequency is open to debate, but it does appear that there is a very strong connection between input frequency and how well we understand a sentence.

Is this a problem? Perhaps not. And I would never question the reliability of an assessment such as the TROG which has been carefully designed and rigorously standardised. But if we’re presenting sentences which children almost never hear, then it does beg the question “what exactly are we testing?” Answers on a postcard please.

Fox, B. A., & Thompson, S. A. (1990). A discourse explanation of the grammar of relative clauses in English conversation. Language, 297–316.

Friedmann, N., & Novogrodsky, R. (2011). Which questions are most difficult to understand?: The comprehension of Wh questions in three subtypes of SLI. Lingua, 121, 367 – 382.

Kidd, E., Brandt, S., Lieven, E., & Tomasello, M. (2007). Object relatives made easy: A cross-linguistic comparison of the constraints influencing young children’s processing of relative clauses. Language and Cognitive Processes, 22(6), 860 – 897.

Nick Riches